STL format, again

From: André Dolenc (Helsinki University of Technology)
To: RP-ML
Date: Monday, December 4, 1995
Subject: STL format, again
Peter H. Gien writes:
 > enlighten us on how many triangles would be needed to represent realistic
 > surfaces, such as grass and hair?
      Peter, it boils down to the size of the model and the amount of accuracy you want. No one is disputing whether you need more geometrical elements using triangles instead of high-order surfaces. Besides, you don't really represent *textures* explicitly in a CAD system, do you? A texture in many CAD systems is a *property* of the surface used when a picture is rendered.
     You will never NEVER represent textures like grass, glass, hair, explicitly in a 3D solid CAD systems; most of the ones used nowadays---ACIS-based in particular---use Breps, and doing so would blow away all your memory, even the xGB in a workstation. For that you would need to use CSG-based reps.
     When you have in your possession an RP machine, you want to build parts starting from anything that resembles a 3D model. Do you have any idea how many CAD systems there are out there? The last time I looked at a catalogue from Silicon Graphics, there were more than 50 CAD systems for mechanical design alone.
     Data transfer is a critical issue. When done via STL, you are putting the responsibility of obtaining a solid were it belongs: the designer. In RP, there will never be one standard for data transfer. Users will and must be able to choose the route which suits their requirements best. You will always find a situation where an STL representation is better than, say, an IGES files, and vice-versa.
     I hear "contours"? Yes, some people will be very happy to send contours, but that determines the surface finish *and* the costs of building a part. How many users of RP have the necessary knowledge of a specific process to do *that*?
Regards, Andre'